In Anneli Pajunen (ed.) Mimesis, Sign and Language Evolution, Publications in General Linguistics 3,University of Turku, Finland.

Mimesis:The “missing link” between signals and symbols

in phylogeny and ontogeny?

Jordan ZlatevDepartment of LinguisticsLund University, SwedenJordan.Zlatev@ling.lu.se

1. Introduction

The purpose of this chapter is to explore the idea expressed in the title: that mimesis, thevoluntary use of one’s body as a representational device, played a crucial role in theorigins of human language and thought, and continues to do so, though possibly to alesser extent, in their development in children.

The basic idea is not new, but has been recently elaborated and defended with respect tohuman phylogeny by Donald (1991, 2001), and applied to child development by Nelson(1996), and (independently) by Zlatev (2001a, 2001b). At the same time, with thestrongly rekindled interest in the evolution of language, similar but competingevolutionary theories have instead attributed a pivotal role to praxis (Corballis 1991),gesture (Armstrong, Stokoe and Wilcox 1995), imitation (Tomasello 1999) and ritual(Deacon 1997). I would suggest that the battle is at least in part terminological, reflectingthe importance of “branding” in modern capitalist culture (Klein 2000), with theoristskeen on emphasizing their own favorite concept at the expense of those of competitors.My choice falls on Donald’s work, since it stands out as the most integrative, and hisconcept of mimesis as most comprehensive. It has been defined as follows:

Mimetic skills or mimesis rests on the ability to produce conscious, self-initiated, representational acts that are intentional but not linguistic.(Donald 1991: 168)

Mimesis is the result of evolving better conscious control over action. Inits purest form, it is epitomized by four uniquely human abilities: mime,imitation, skill, and gesture. (Donald 2001: 263)

In the latest work, these four mimetic abilities are placed in a tentative evolutionaryhierarchy. Mime is considered simplest since it (only) involves a reenactment of anexternal event. Imitation is more complex because it requires grasping the purpose

(intention) behind the modeled event, as also emphasized by Tomasello (1999). Skill, isclosely based on the previous two, since it involves the rehearsal (self-imitation) of aseries of hierarchically organized mimetic acts, until some ideal standard is reached.Finally, gestures are explicitly communicative, often elaborately sequenced bodymovements, requiring a shared understanding of their meaning for communication tosucceed.

This latter definition of gesture is considerably narrower than that of Armstrong et al.(1995), who define it as “a functional unit, an equivalence call of coordinated movementsthat achieve some end” (ibid: 46). This includes just about anything from a goal-directedleap to the “visual gestures” of Sign1 languages, and the “invisible vocal gestures” ofspoken languages, and therefore represents a concept so general as to be practicallyvacuous. However, it does serve the gradualist theory of language origins espoused byArmstrong et al. (1995), according to whom the differences between “ape gestures” andSign language (or vocal language, for that part) seem to be a matter of greater complexityand conventionalization. If this were indeed the case, then there is simply no gap for any“missing link” to fill.

But there are some compelling reasons to believe that there is one. Donald clearlydistinguishes the “event representations” of apes from the language-basedrepresentations of human culture. So does Deacon (1997) referring to the cognitive andcommunicative categories of animals (mistakenly) with the Peircean terms “icon” and“index” (Pierce 1931-35). Likewise, Tomasello (1999) distinguishes between “gesturalsignaling” and gestures proper, arguing that apes (in the wild) are only capable of the firstsince they lack an understanding of others as psychological beings, which constitutes arequirement for intentional communication (Grice 1957). Indeed, the differences betweenanimal signals and human symbols (Vygotsky 1978, Deacon 1997, Tomasello 1999,Sinha 2001, this volume), some of which are summarized in Table 1, are so considerablethat it is almost impossible to envision a single transition from the first to the second.This has provided fuel for otherwise evolutionary anomalous, strongly nativist,suggestions of a sudden grand mutation, creating language almost ex nihilo (Bickerton1990, Piattelli-Palmarini 1989). It is also characteristic that Deacon’s attempt to explainthe transition to symbolic reference with the onset of marriage ritual is where his theoryis least convincing (cf. Armstrong 1998).

1 I adopt the convention of spelling “Sign” with a capital letter, when it used to refer to the manual-brachiallanguage of the deaf.

Table 1. Some differences between signals and symbols in terms of relevant cognitive characteristics,making the two categorically distinct (cf. Vygotsky 1978, Deacon 1997, Sinha 2001, this volume). Notethat there can be quantitative differences within each category.

Signals SymbolsExamples: bee “dance”,

bird song,vervet monkey calls

Kanzi’s “lexigrams” ,words,mathematical symbols

Degree of learning None (genetically determined)or limited

Extensive

Conscious control overproduction

None, or highly limited High (in normal cases)

Interpretation (Relatively) fixed response Flexible, “negotiable”Contextuality Tied to a particular context

(stimulus setting)Flexible, relatively independentfrom specific context

Systematicity(interrelatedness andcombinatorics)

None, or very limited High

A chief advantage of Donald’s theory is that it has suggested how the bridge betweenanimal and human communication and cognition can be built not by a miraculous“chance mutation”, but by mimesis which gave rise to the first public representations andthus serving as the basis for a culture of shared scripts, ritual, pedagogy and gesturalcommunication. On their part these cultural skills prepared the transition to the oral-linguistic culture of Homo sapiens.

It is this hypothetical role of a “missing-link” played by mimesis that will be the topic ofthis chapter. In Section 2, I will review some of the evidence for a “mimetic stage” inevolution presented by Donald, but also by others. Since ontogenetic development can beexpected to resemble in broad lines – though not to “recapitulate” – phylogeny due to itsepigenetic character, we should expect mimesis to be strongly present there too, andSection 3 will discuss some relevant phenomena in child development. While these twosections deal with the hypothesis empirically, Section 4 will briefly address the questionfrom a more conceptual perspective: given the consensual, intersubjective nature oflanguage, and the nature of our bodies and brains, could mimesis even be a necessaryprecursor to language? Finally, Section 5 summarizes the conclusions of the study, andpoints out implications.

Before proceeding, I wish to call attention to a typographical convention that will beadhered to: Since the text is full of many theoretically loaded terms, some of which havedifferent everyday meanings, for the sake of clarity I will continue to place such terms inboldface, and whenever possible will add a definition. When a definition is absent thereader may regard the boldface as a “warning sign”, and if unsure to consult the sources.

2. Evidence for a “mimetic stage” in hominid evolution

2.1. Two major transitions in biological and cultural evolution

There is considerable agreement among paleontologists and physical anthropologists thatwith Homo habilis/erectus some 2 million years BP (before present) there was both acultural and a neuroanatomical “jump” in hominid evolution. The transition had begunearlier, with the branching off from the chimpanzee line about 5-6 million BP, and gainedmomentum with the different species of the genus Australopithecus (who were allbipedal) some 4,000,000 years BP. However, the biology and culture of Homo erectuswas the first that showed a qualitative break with the animal world. Their brain/body ratiowas twice that of apes – though still 70% that of modern humans. They constructed fairlyelaborate tools, migrated out of Africa and colonized most of the world, using fire andbase camps. At the same time, their vocal tracts were still similar to those of apes, andthere is no evidence that they possessed any kinds of vocal language. They did have ahomologue to Broca’s area – implicated in the use of language in modern humans – inthe left hemisphere, but this has also been shown to be involved in praxis (Corballis1991) and imitation (Rizzolatti and Arbib 1998), as explained below. As summarized byDonald (2001: 267):

For over two million years hominids shared food, disseminated skills byimitation, created mini-industries for activities such as tool manufacture,and held group behavior to very stable long-term patterns, such as thecontinuous habituation of a site and the maintenance of fire. These thingsare not easy to achieve, and there is no good evolutionary rationale forattributing them to language.

But this long period of relatively continuous development was interrupted by a secondtransition some 200,000 - 150,000 years ago with the advent of Homo sapiens: brain size,especially due to prefrontal cortex expansion, radically increased and reached modernproportions and the vocal tract attained its present form. This coincided with a culturalexplosion and a new exodus “out of Africa” leading to a colonization of the world thatwas to be total, wiping out any predecessors such as the Neanderthals in the process. Asmany have argued (e.g. Lieberman 1991), there are good reasons to think that a crucialadvantage in Homo sapiens compared to other species and predecessors was thepossession of a high-speed, articulate vocal language. But in that case, what was it thatserved the basis for Homo erectus culture? Mimesis, rather than either spoken languageor Sign language, stands out as a likely candidate.

2.2. “Mimetic vestiges” in modern cultures and brains

A second sort of evidence appealed to by Donald is based on the metaphor of “vestiges”– traces of earlier stages of evolution in modern human culture and communicativebehavior. On an intuitive level, our fascination with many forms of art (dance, theatre,pantomime), games (“follow the leader”, charades) and athletics testifies to the centrality

of mimesis in our daily life – in the form of mime, imitation, motor skill, or combinationsof these.

But perhaps the clearest evidence for a mimetic “substratum” underlying human verbalcommunication is the ubiquity of gesture, especially of the type called by Kednon (1988)gesticulation and by McNeill (1992) speech-associated gestures (cf. Gullberg 1998).Such gestures are usually classified as iconic and metaphoric (which depict concrete orabstract concepts respectively) or rhythmic (“batons” or “beats”). What is curious is thatsuch gesturing is largely performed unconsciously, and at least in part unintentionally:even blind people gesture when speaking. The fact that it is closely synchronized withspeech has lead McNeil to claim that the two communication systems share a commonunderlying representation, and therefore that gesture is no less “verbal” than speech:“Gesticulation and language are seen as equipotent reflections of thought with differentoutput channels. They are interdependent and one is not a translation of the other”, assummarized by Gullberg (1998: 62). This interdependence has been further supported byrecent evidence for gesture differences between Spanish and English speakers whichcorrelate with semantic characteristics of the respective language (McNeill 2000).

On the other hand, McNeil (1992) himself states that speech-associated gestures lackhierarchical structure and semantic compositionality – two characteristics usuallyconsidered defining of linguistic representation – which places in doubt the“equipotentiality” of language and gesture. Evidence from aphasia shows that while lossof speech is often accompanied by lack of appropriate gesturing, this is not so in all cases,in particular in individuals with global aphasia, completely unable to produce andunderstand speech (even as an internal monologue), but maintaining the ability tocommunicate with iconic gestures of the type that are usually “speech-associated”, aswell as through miming, pointing and by using standardized, culture-specific emblems.2

All this supports the thesis “that mimetic representation is the more fundamental level ofrepresentation” (Donald 1991: 224), separate from, but in normal conditions interactingwith, and under the influence of the evolutionarily more modern linguistic system.

A fairly clear demonstration of the fundamental though “sub-verbal” character ofmimetic representation can be seen in the gestural communication of deaf people who arenot exposed to Sign. Such individuals, e.g. deaf children of speaking parents, interactwith their hearing environments or with other “Signless” deaf people through a form ofpantomime that has come to be called homesign, studied extensively by Goldin-Meadowand her colleagues. It has been established that over time, processes of simplification andstandardization make the communication system more arbitrary and compositional, i.e.less mimetic and more linguistic (Morford, Singleton, Goldin-Meadow 1995):

After a language has undergone several generations of change, very littleattention is paid to the iconicity of symbols by the language user at all.Thus, it is only when systematicity is given priority over iconicity thatsymbols are transformed into language. (ibid: 330)

2 Such as the “OK” sign performed by touching the tops of index finger and thumb.

On the one hand, this can be said to show the difference between mimesis and language,with the transition between the two providing a kind of “recapitulation” of evolution. Onthe other hand, the readiness with which speech-impaired children acquire Sign language,or even “create” one in the absence of a model, has lead some researchers to propose thatnot only mimesis, but a full-blown sign language preceded spoken language in evolution(Armstrong et al. 1995, Stokoe 2001). This is possible, but by no means necessary. If weassume that mimesis and speech constitute interacting systems (as the evidence suggests)we can see how a brain adapted for the acquisition of symbols represented in speech,such as the brain of Homo sapiens, would also be one adapted for the acquisition of Signthrough the “feedback” projections of the more modern system (language) over the moreancient one (mimesis), thus recruiting it for symbolic communication in the context ofdeafness. This is also consistent with the evidence outlined in Section 2.1 that Homoerectus hardly had access to a full language, either spoken or Sign.

2.3. Manual control and bipedalism

One could argue that the evidence reviewed above is consistent with the possibility thatvocal language developed in parallel with and independently from mimesis and gesture,linking with it only after prolonged evolution (and never completely, as the data fromaphasia suggest). One such scenario is suggested by Deacon (1997) according to whomthe first symbols arose through pair-boding ritual as early as the genus Australopithecusand started a chain of brain-culture co-evolution – all this with little or no role formimesis. This scenario is also consistent with evidence for a gradual evolution of thevocal tract from Homo erectus to anatomically modern Homo sapiens (Lieberman 1991).

However, one difficulty in assuming that the first “symbols” were vocal is that in thatcase they would have originated from the primate call system, which is far less subject toconscious control than the limbs. This is illustrated by the anecdote told by Deacon(1997) of a chimpanzee who has found food in a secluded place and attempts to mufflehis involuntarily produced food-call with its hand, in order to be able to enjoy the foodalone. Recent neurobiological evidence shows that it is a region in cingulate cortex thatcontrols primate calls and that this region is not at all homologous (i.e. having the sameorigin) to Broca’s area in modern humans, which as well-known (though notuncontested) plays a key role for the grammatical organization of language. Broca’s areahas rather been argued to be homologous to areas in the pre-motor cortex involved inmanual-brachial (hand-arm) control and action recognition (Rizzolatti and Arbib 1998).

With the onset of bipedalism in Australopithecus the degree of manual control andproficiency would have increased even more. While there are many other possibleevolutionary pressures to explain the transition to bipedalism – e.g. change of ecologicalniche, childrearing and tool-making – mimetic gestures could clearly have been a factorfor establishing the transition. It is quite possible to imagine a scenario in which anexisting manual communication system eventually “recruited” vocalization, bringing itunder voluntary control and providing the selection pressures for lowering the larynx.The advantages for this latter change must have been quite strong to counterbalance the

negative effects of more likely choking, which would only have been the case if there hadbeen language-like communication prior to anatomically modern Homo sapiens, which isstrong evidence against a relatively late emergence of all linguistic skills (Johansson2002).

2.4. Co-lateralization for language and “praxis”

As well-known, language skills are “lateralized” in about 90% of all people, meaning thatmost of the crucial (though not all!) areas for language production and comprehension arelocated in the left hemisphere, above all the classical two: Broca’s area in the leftfrontal pre-motor cortex, and Wernike’s area in the left temporal gyrus. A similarproportion of the population is “right-handed”, in the sense that precise and complexmotor activities are better performed with the right hand (and foot), both of which areunder the control of the left hemisphere. Among right-handed people, the co-relation withleft-hemisphere specialization for language is nearly perfect. With left-handers, themajority still have left-hemisphere dominance for language, but there is a sizableminority who don’t. What is the reason for this strong, if not perfect correlation betweenlanguage dominance and handedness? Corballis (1991) has proposed that not justlanguage, but all behavioral systems involved in assembling complex sequences ofactions (including visual routines) are controlled by the left hemisphere, which isendowed with a specific cognitive adaptation, termed generative praxis. The claimedorigin of this capacity is the refinement of manual skill, and especially tool-making inHomo erectus some 2 million years BP. Greenfield (1991) is even more specific inclaiming that Broca’s area itself is implicated as the cerebral cortical basis for thehierarchical organization of language and manual object combination, presentingevidence that the two abilities develop simultaneously in human ontogeny.

All this is, of course, consistent with a central role for mimesis, since motor skill isaccording to Donald a mimetic ability. But while mimesis in general implies the capacityfor representation, praxis does not. And since representation is a central characteristic oflanguage, it remains unclear why language should “piggyback” on, or develop fromstructures for praxic skill. What is necessary is to show that the “proto-Broca” areas wereinvolved not just in the control of action sequencing, but in semiotic activity, i.e. the useof signs – in communication, thought or both.

The missing piece of the puzzle could possibly be provided by the recent discovery of the“mirror-neuron” area F5 in the pre-motor cortex of the monkey brain, which mediatesbetween observed actions and performed action, and which has been persuasively arguedby to be homologous with Broca’s area in modern humans, as suggested below.

2.5. “Mirror neurons”, iconicity and parity

Rizzolatti and Arbib (1998) and Arbib (in press) sketch the following scenario for therole of the mirror neuron system in the “proto-Broca” area in the development of

mimesis. Initially it was used for action recognition, as it still is in monkeys. In the greatapes, and in our common ancestor some 6 million years BP, it was extended for thepurpose of imitation of goal-directed actions. A crucial step was realized in hominids,who began using the system for the production, perception and imitation of actionswithout the involved objects, in other words for pantomime. This according to Arbib (inpress), would have been the first form of explicit representation. For example if hominidA performs a grasping motion towards an imaginary fruit, hominid B would recognize hisactions and would be likely to understand it as an expression of a certain intention(provided at least a rudimentary “theory of mind”). Notice that this scenario is differentfrom Donald’s (2001) mimetic hierarchy (mime > imitation > skill > gesture), by placing(panto)mime, which is explicitly representational, after imitation/skill, which are not.This reordering is quite plausible from a Vygotskian “Outside-Inside” perspective,according to which “highter mental functioning” first emerges between people incommunication, before it can mediate between the subject and the physical environment,and even later within the individual’s own mind.

An advantage of the above hypothesis is that it does not require a major “insight” that thebody could be used as a representational tool from some imaginary early hominid formimesis to get off the ground. Since learning by imitation, and a social life dependent onit, would have already been quite established, at the latest by the time of Homo erectus,all that would have been necessary is for a “teacher” occasionally to perform a particularaction without the necessary object, to make this a highly iconic sign of that action. Sincethe mirror system provides a basis for parity – an externally observed and a self-produced action are in some sense categorized as equivalent – the sign, as well as the fullaction it signifies will be “shared” between sender and receiver.

Another advantage is that in this way we can see how the “germ” of grammar would havebeen inherent in the first communicative acts, which could have consisted of (eventuallystylized) actions with “slots” for the participant roles: the Actor, the Patient, theInstrument etc. With standardization, these different parts would have been furtherdifferentiated from the action itself, which is conceptually much simpler than imagininghow “things”, “actions” and “properties” could be segmented and named individually andcombined afterwards. Armstrong et al. (1995) point this out as an argument for thegestural origin of language:

Not only visible gestures … often resemble what they signify, they arealso likely to have originated … [through the discovery] that certainactions observed were, and others could be, mimicked with manual-brachial actions. … [N]othing so far seems to have as much potential asvisible actions involving an actor and its action for guiding hominids tothe act of symbolizing relations as well as things. (ibid: 22-23)

The hypothesis is thus that grammatical structure was initially highly iconical, and thisrhymes with the observation that it largely still is so in Sign languages (Armstrong et al.1995) and to a lesser extent even in spoken languages (Haiman 1985, Itkonen, thisvolume). Degree of iconicity, however, should not be taken as a simple indication of how

ancient a communication system – and thus that full Sign language intervened betweenmimesis and spoken language.

2.6. Summary and unresolved issues

Even the cursory review of some of the evidence for the role of mimesis in humanevolution presented above shows that mimesis has many of the properties necessary forbridging the gap between animal signal systems and human symbolic communication,and particularly language:

• (partial) generativity• intentional communication• public representation• parity• iconicity

On the one hand it is readily imaginable how the early hominids developed such asystem, and on the other, how it was further transformed towards greater symbolicity (cf.Section 1). While it is not impossible that this later development went through a “stage”of more or less modern-like Sign language, it was pointed out that this inference is by nomeans necessary, and probably less likely than a scenario in which speech began to be“recruited” for disambiguating between gestures fairly early, and with time and processof “Baldwinian evolution” to take over as the main vehicle of communication,interacting with but separate from mimesis. If this was so, then the transition fromgestural to vocal communication was not only one of modality change, but one ofqualitative semiotic change, where a mimetic-iconic system is supplemented by asymbolic one. In this case the motivation for the transition would not only have been“external”: e.g. freeing the hands and visual attention for other means, communicatingover barriers and in the dark, better pedagody in tool-making (cf. Givon 1998, Corballis2002), but also “internal”, since the vocal medium would automatically have introducedgreater arbitrariness (Saussure 1916, Morford et al. 1995) in the mapping betweenexpression and meaning and greater differentiation, thus pushing the development of“protolanguage” to higher symbolicity.

The discussion on whether mimesis first developed into a manual-visual language, orwhether it served as a stepping stone to oral language directly parallels the disagreementon whether gesture and language constitute a single system (McNeill 1992), or rather aredifferent evolutionary layers of the “hybrid mind” (Donald 1991). If gesticulation andlanguage do differ in terms of hierarchical structure and compositionality, as McNeillhimself claims, and if the evolution of oral language transformed mimesis into somethingqualitatively different: a conventional symbolic system, then despite claims to thecontrary, gesticulation is more mimetic and less “verbal” than speech (and Sign).

On the other hand, the major drawback of the concept of mimesis, as used by Donald isthat it is too inclusive. As shown there is quite a difference between skill, imitation, mimeand gesture (and even more so, when e.g. ritual and dance are included) to regard it as a

single “stage” in evolution. More research is necessary in order to determine the mostlikely progression, and some of the relevant evidence may come for child development,to which we turn in the next section.

3. Evidence for mimesis in ontogenetic development

In contrast to both nativist and empiricist approaches to human cognitive development,the paradigm of epigenesis (Piaget 1971, Waddington 1975, Sinha 1988, Zlatev 1997)proposes that ontogeny can be explained as an ordered succession of developmentalstages toward higher complexity that emerges through interaction with the environment,influenced by but not determined by the genes. It is a strong alternative since it isscientifically parsimonious (e.g. no commitment to an innate Universal Grammar, cf.Haukioja this volume), evolutionary plausible (Müller 1996), descriptively adequate(Nelson 1996), and theoretically attractive in providing a way to integrate the “natural”and “cultural” lines of human development (Vygotsky 1978, Tomasello 1999).

Another advantage, less often noticed, is that epigenesis also offers a basis for integratingstudies of evolution and ontogeny, since it provides a long-missing rationale forHaeckel’s (1874) famous “biological law of recapitulation”, albeit in a weaker form:Since evolutionary more ancient cognitive structures (in general) require less experiencein order to mature than evolutionary more modern structures, the early stages ofontogenesis will (in general) be dominated by evolutionary more ancient mechanisms,and the later stages by more modern mechanisms. In some cases the “lower” stages mayeven serve as an epigenetic prerequisite, i.e. a necessary step, for reaching the “higher”functions.

The question to ask in the present context is if mimesis – in its different manifestations –may be said to function as a “stage” in human ontogenetic development, and if it playsthe role of an epigenetic prerequisite for the acquisition of symbols in general, andlanguage in particular. Such a mediating role of mimesis in human development has beenproposed independently by Nelson (1996) and Zlatev (2001a, 2001b), but with ratherstrikingly different conclusions: While according to Nelson mimetic representationdominates human cognition in early childhood, approximately from 1,5 to 4 years, in myapplication of Donald’s evolutionary model to ontogeny, I discerned an analogous“mimetic stage” ranging between the landmark developments of the “9 monthrevolution” (Tomasello 1999, cf. below) and the vocabulary and grammar spurts towardsthe end of the second year. This discrepancy calls for an explanation. After a review ofsome of the relevant evidence for the nature and timing of mimetic skills during the firstyears of life, I will try to determine which (if any) of the proposals is most consistent withthe facts, and with the concept of mimesis itself.

3.1. Innate predispositions for imitation

Perhaps the most widely publicized results on infant imitation during the last 25 yearshave been those of Metzoff and Moore (1977, 1983, 1995) which demonstrated that new-born babies are capable of imitating simple movements involving mouth-opening,

tongue-protrusion and lip-protrusion, thereby showing that the human capacity to imitateis at least partially innate. The capacity to perform such “blind imitation” in the absenceof a unitary sensory metric in which to compare one’s own action with that of another(which, notice, exists for oral-vocal imitation) is really quite amazing, and it isunsurprising that it is still sometimes contested. In terms of the subsequent findings of a“mirror neuron system” mentioned in Section 2, this inborn capacity can possibly beunderstood as a partial “pre-wiring” resulting in a mapping between the neuralrepresentations of the baby’s own body movements and visually induced representationsof the movements of another human being, though how this can be realized in specificterms is still anybody’s guess. Granting the validity of the phenomenon, would it bejustified to claim that mimesis is present from birth? Hardly, unless one is also preparedto describe the neonate’s imitations as “volitional acts of representation”, which evenMeltzoff and Moore refrain from, given these movements’ limited and stereotypicalcharacter.

But in a more recent study Meltzoff and Moore (1995) report imitation of effortful, notnaturally occurring behavior such as moving the tongue from side to side from as early as6 weeks, and argue that infant imitation has the following features that distinguish it frommimicry: intentionality (ability to distinguish the goal state from the actual performanceand correct the latter in terms of the former), selectivity (imitating different aspects of themodel), creativity (original ways of reaching the same goal) and volitional control(ability to differ imitation, possibly indefinitely). This is quite a tall order, and if grantedwould imply the existence of mimetic skill soon after birth. But while the reported datastrongly suggest intentionality (in the sense of “goal-directedness”) and some creativity,the other features are quite probably attributed due to an exceptionally “richinterpretation”. For true volition the 6-week old infant must possess self-consciousness,and there is no evidence to support this. Furthermore, mimesis possesses one morecrucial feature, which would naturally follow with the above four, namelyrepresentation, the ability to distinguish between “signifier” (one’s own body) and“signified” (another’s action or some other event or object), and to realize that the firststands for the latter – a capacity that is even less imaginable in the new-born, as alsoadmitted by Meltzoff and Gopnik (1993).

The most likely conclusion, consistent with the epigenetic standpoint, is therefore thatMeltzoff and Moore’s studies of infant imitation have uncovered an adaptation formimesis, but that these innate skills are only “bootstraps” – in the manner of suckingreflexes but on a higher level of complexity – which require adequate social interaction inorder to lead to the development of imitation proper, possibly along with self-consciousness and reference, as suggested below.

3.2. Imitation, intersubjectivity and representation

As emphasized by Donald and Tomasello, true imitation can be distinguished frombehaviors such as mimicking and stimulus enhancement by the fact that it involves themodeling of goal-directed behavior, which requires at least some understanding of the

model as an intentional agent. Tomasello presents much evidence that such anunderstanding develops first around 9 months, summarizing the transition with thememorable phrase “9-month revolution” and describing it as follows:

At about 9 months of age, infants begin to behave in a number of waysthat demonstrate their growing awareness of how other persons work aspsychological beings. They look where adults are looking (joint attention),they look to see how adults are feeling toward a novel person or object(social referencing), and they do what adults are doing with a novel object(imitation learning). ... Infants also at this time first direct intentionalcommunicative gestures to adults, indicating an expectation that adults arecausal agents who can make things happen. All of these behaviorsindicate a kind of social-cognitive revolution: At 9 months of age infantsbegin to understand that other people perceive the world and haveintentions and feelings towards it; they begin to understand them asintentional agents. (Tomasello 1995: 175)

On the other hand, the well-known infant developmentalist Threvarthen (1992) hasargued persuasively for a kind of primordial intersubjectivity (sharing of experience)between infant and caregiver present almost from birth. But also Threvarthenacknowledges the transition emphasized by Tomasello, referring to it as the attainment of“secondary intersubjectivity”. As was the case with imitation, joint attention has itspredecessors (Butterworth and Jarrett 1991), but it is only toward the end of the first yearthat children begin to participate in a “referential triangle” with an adult and an externalobject or event, where both are aware that they are sharing attentional focus. A theoreticalgeneralization of these findings can be stated if we assume, largely following Tomasello(1999), that the underlying development in all these cases is the understanding ofintentionality, motivating the child to do what the adult does for the reason of achievingthe same goal as the adult: paying attention to “the same event”, reacting in “the sameway” to a new person, performing “the same action”.

However, it need not be as suggested by Tomasello that this is done (solely) in the Inside-Out manner of “mental simulation”: the child first gains an understanding of its ownintentions, and then projects these to other beings who look and behave similarly. Thecausality can run in both directions, as suggested by Baldwin (1913): “My sense ofmyself grows by imitation of you and my sense of yourself grows in terms of myself”, oras I have tried to elaborate (Zlatev 2000), in the form of a “loop” involving (1) increasedobjectification of one’s body, (2) increased volitional control, (3) increased “theory ofmind” for others and oneself, grounded in the “bridge” of primary intersubjectivityprovided to a large extent by the innate imitation capacity (Meltzoff and Gopnik 1993).

Though it still remains in question how it is achieved, it is quite clear that one of thecrucial features of mimesis – awareness of one’s body – has already emerged by the endof the first year. What about representation? In Piaget’s (1952, 1953) “classic” theory ofthe development of representational ability (which unlike many modern uses of the term“representation” implies accessibility to consciousness), this is achieved at a later age

towards the end of the “sensorimotor period”, approximately around 18 months. Bypresent data, this seems to be an underestimation of the child, conditioned both byPiaget’s limited evidence and by his general model.

But an important similarity with more modern developmental theories (such as thosementioned earlier) is that Piaget too considered imitation as pivotal for the origin of thefirst “symbols”, iconic mental representations schematizing instances of sensorimotoractivity. It would be much more consistent with current terminology to refer to these asmimetic schemas. The developmental scenario proposed is furthermore surprisinglyreminiscent of Vygotskyan internalization: first the child externally (overtly) imitatesthe actions of another person, followed by imitation of oneself in skill development,followed by the internal (covert) imitation of an inner model in imagistic thinking. Bothexternal and internal imitation can be not only of concrete actions, but of other eventsperceived as similar, thus leading to mime. So despite his strongly individualisticapproach to the origin of representation, Piaget also attributed a central role to mimesis inthe transition to “signs” and language, and his developmental trajectory: imitation > skill> mime > mimetic schemas is consistent with the modern evidence, as well as theevolutionary progression suggested in Section 2.

3.3. Mimetic gestures and “baby signs”

Even if the child would be capable of developing mimetic schemas for representationalthought without engaging in interpersonal communication, as implied by Piaget – whichis doubtful (but empirically untestable) – it is clear that being embedded in a culturewhere such mimetic schemas are used referentially, i.e. in interpersonal communication,would boost the child’s mimetic skills even more. As mentioned earlier, infants of 9+months begin using mimesis for intentional communication, where the most salientcommunicative gesture is pointing, and a few others such as waving “bye-bye”, “come”or “no” which are acquired spontaneously. But Acredolo and Goodwin (1996, 2000) haveshown that from the age of 9 months (again!) children are capable of learning many so-called “baby signs”, motivated gestures, associated with a conventional meaning byparent and child, enabling the two to communicate with shared concepts even before thechildren can acquire language:

With encouragement from parents, babies can learn to associate dozensand dozens of gestures with specific things-like flapping arms for bird,smacking lips for fish, blowing for hot, or even patting the chest for afraid.(Acredolo and Goodwin 2000: 84)

The exact shape of the signs of course does not matter, but it was shown thatmotivatedness helps children acquire them, i.e. the system is one of (mostly) iconicgestures. Acredolo and Goodwin also present evidence that children taught “baby signs”are faster to acquire vocal language and even have a higher IQ than peers at the age of 8.This has raised both controversies and something of an industry for teaching baby-signsin middle-class families, but from an epigenetic perspective it is little surprising that

culturally stimulating a “natural” capacity for developing representation and referencewould have beneficial effects on development. Furthermore, it constitutes evidence thatmimetic and subsequent linguistic representations are related, as held by Piaget (1952),rather than constituting separate “layers” in the “hybrid mind”, as sometimes suggestedby Donald (2001).

3.4. “Symbolic play”, differentiation and conventionalization

As noted, Piaget (1952, 1953) considered the “symbolic function”, or representationalability, developing during the second year of life to be essentially an individualachievement, something that has been strongly challenged by pointing out the strongearly influence of social mediation (Vygotsky 1978, Nelson 1996, Tomasello 1999) onthe infant’s mind. But even Piaget held that the “symbols”, or rather the mimeticschemas, of the late sensorimotor period were semiotically deficient in at least two ways:they were not fully differentiated from their referents, and they were idiosyncratic.Therefore they were not completely adequate both for either thinking and forcommunication.

It is characteristic that at this stage of development, referred to by Piaget as the “pre-operational period” children become increasingly involved in symbolic play, orpretence, both with objects and with other children. Furthermore, these interactionsdisplay a rising degree of complexity and abstractness:

Between 18 months and 2 years of age, make-believe becomes moredetached from the real-life conditions associated with it. In early pretencetoddlers use only realistic objects – for example, a toy telephone to talkinto or a cup to drink from. Around age 2, use of less realistic toys, such asa block for a toy receiver, becomes more frequent. Sometime during thethird year, children can imagine objects and events without any directsupport…. (Berk and Winsler 1995: 55)

Thus, symbolic play can be seen as a “mini-laboratory” in which children experimentwith representations, and in the process both learn to distinguish these from reality, howto make them more abstract, and (if necessary) to internalize them. Since these arerepresentations of activities, we can safely conclude that they are wholly, or largely,mimetic. What is the relationship between these and language? Many studies have showna correlation between stages of the development of symbolic play and early linguisticdevelopment (cf. Nelson 1996). The causal relationship in rather unclear, but it is naturalto suppose bi-directional interaction, since in both cases (symbolic play and language)similar processes of differentiation and conventionalization of the child’s representationsare played out.

Another similarity between the structure of symbolic play and language, pointed out byVygotsky (1978), is that both require the mastery of social, normative rules which needto be internalized, i.e. the child not only acts in certain way, but understand that it should

act in that way. While it has been questioned whether children’s early language displayssuch awareness of normativity, it is clear that when children are capable of playing agame with roles and symbolic objects (either alone or together with others) they must be,in some sense, following normative rules: “Whenever there is an imaginary situation,there are rules.” (ibid: 95). Therefore, it is probable that mimesis is (again) leading theway, and language is following. This can also be observed in older children, when thelevel of linguistic competence is not sufficient for a specific function. For example,Evans & Rubin (1979) show that kindergarten children use gestures instead of speech inexplaining game rules, whereas the gestures of five- to ten-year-olds are much moreredundant in relation to their verbal expressions (reported by Gullberg 1998:75).

3.5. Pedagogy: ZPD and “scaffolding”

Not only language, but also pedagogy has been convincingly argued to be a distinctlyhuman trait (Premack 1987, Donald 1991, Tomasello 1999) and it is not difficult to see aconnection even though language is not literally “taught” to children. An influentialconcept in explaining the way in which pedagogy empowers an individual’s cognitivedevelopment is that of the Zone of Proximal Development (ZPD) defined by Vygotskyas:

…the distance between the actual developmental level as determined byindependent problem solving and the level of potential development asdetermined through problem solving under adult guidance or incollaboration with more capable peers. (1978: 26)

As similar idea is expressed by the concept of “scaffolding” provided by the adult insupport to the developing child, and gradually “removed” as the child’s competenceincreases. A problem with this metaphor is, though, that it is too static to do justice to thehighly dynamic and adaptive interaction between parent and child, described by Berk andWinsler (1995) as follows:

A useful analogy for the complex interaction that occurs duringscaffolding is an intricate dance between a teacher (the adult) and a pupil(the child), with the child leading and the adult following for instructionalpurposes. (ibid: 29)

The alternative figure of a dance highlights two different ways is which theZPD/scaffolding phenomenon may be said to involve mimesis. The first and more literalis when a manual or bodily skill is being transmitted, requiring imitation learning. Thesecond is more general and concerns:

…the ability of the adult to sense what the child can, and can not learn – inother words to judge the ZPD. This skill is also mimetic, in as much as itinvolves developing a “model” of the child’s current knowledge. (Donald1991: 177)

While one might accuse Donald of overextending the concept of mimesis here – since itis not clear how this involves the employment of representational bodily skills – thesuggestion is worth looking into. The emphasis on “sensing”, rather than “theorizingabout” the abilities of the child is more phenomenologically true to the experience of day-to-day pedagogy, and could help explain its highly dynamic, context-sensitive character.A hypothesis along these lines would be that the neural structures involved in imitation,“mentalizing” (Firth 1989) and “ZPD-sensing” would significantly overlap.

3.6. Summary and conclusions

In this section, I reviewed evidence for a number of related phenomena in childdevelopment, showing both a partially innate predisposition and an important mediatingrole for mimesis:

• infant imitation• the rise of intersubjectivity• the development of mimetic schemas3

• iconic “baby signs”• symbolic play• ZPD and scaffolding

Their role is “mediating” in the sense that they constitute epigenetic prerequisites,“stepping stones” in development, without which further development would be if notblocked, at least seriously compromised. Though some of the relationships are stillconjectural, it seems likely that without imitation (capacity), neither intersubjectivity northe representational capacity evident in mimetic schemas would develop. On their side,mimetic schemas along with symbolic play are instrumental to making individualrepresentations more conventional and normative, and to the meta-representationalawareness necessary for language. The teaching of “baby-signs” is of course notnecessary, but it shows an ability to understand iconic representations from as early as 9months (and seems to have a positive effect on the development of language). Finally,pedagogy involves mimesis through imitation and possibly through the on-line“modeling” of the pupil’s abilities.

But is there conclusive evidence for a mimetic “stage” in ontogeny? The discrepancybetween the time periods assigned by Nelson (1996) and Zlatev (2001a) for such a stageshow that the whole concept may be problematic. Both authors define the type of stagethough “a particular type of representation … dominate[ing] during a particular period ofontogenetic development” (Nelson 1996: 86). But there is a great deal of arbitrariness indeciding which representation “dominates”. The evidence reviewed in this section showsthat while Nelson has underestimated early mimesis, at least from 9 months onwards, inmy previous work, I had simplistically assumed that when language “gets off the ground”

3 Corresponding to the concept of “symbols” (Piaget 1962), and possibly that of “image-schemas”(Johnson 1987, Lakoff 1987).

around the age of 2, mimesis will automatically be assigned a secondary role. The simplecompromise would be to merge the two proposals and extend the mimetic stage from 9months to 4 years. But Nelson offers no hard evidence that language is used only“pragmatically”, rather than representationally prior to the age of 4. The early language-dependency of gesture (McNeill 2000) rather suggests the opposite.

The “stage” concept in ontogenesis has in recent years come under increasing critique forsimilar reasons. An alternative metaphor proposed by one such critical account is that ofSiegler (1996) – children’s thinking develops in the form of partially overlapping“waves”. This concept may do more justice to the evidence presented in this section, aswell as to the concept of mimesis itself. The point is to re-conceptualize the epigeneticrole of mimesis not as a “stage” but rather as waves of mimesis, preceding waves ofsymbolization in a life-long perspective. Instead of mimesis “giving way” tosymbolization, it rather starts before it, and makes it possible by providing varioussupporting structures. When symbolization, in particular language, rises – the two start adance, with mimesis serving as a backdrop, and taking precedence whenever necessary,such as when a new activity is encountered (kindergarten games or a new language) orwhen language is disrupted (aphasia). Such a view saves us from the need to determinean arbitrary point in time when mimesis is no longer “dominant”.

4. Mimesis: a necessary precursor to language?

After having discussed whether mimesis may have served as the “missing link” betweenape and human cognition and communication in evolution, and whether it helps the pre-linguistic child enter the symbolic world surrounding her in the previous two sections, Iwish to consider briefly whether language may be not only contingently dependent onmimesis, but necessarily so. By this I mean a kind of combination of conceptual andnatural necessity as will be exemplified below. At least three different aspects of mimesiscan be (and have been) presented as candidates for playing such a role: its role for skilldevelopment, for iconic representation and for intersubjectivity.

4.1. Mimesis as skill

Considering the broad definition of mimesis involving the “volitional control of thebody” and the fairly uncontroversial claims that (a) no complex skill can be learnedwithout such control and (b) language consists at least in part of such complex skills, itfollows logically that mimesis is a prerequisite for the learning of language, either asspeech, Sign or writing. This rather obvious conclusion is often overlooked in theories of(child) language acquisition, and is therefore correctly pointed out by Donald:

Mimesis is logically prior to language, because without it, we cannotrehearse or refine any skill, let alone one so complex as speech orlanguage. (Donald 2001: 268)

But the general way in which mimesis-as-skill serves as a prerequisite for learning anycomplex motoric skill makes it rather uninteresting as far as the specificity of language isconcerned. There is after all, nothing semiotic per se in motoric skills, while the crucialproperty of mimesis is its representational character. Furthermore, it was suggested thatboth in phylogeny (cf. 2.5) and ontogeny (3.2) the self-consciousness necessary forvolitional body control may itself be dependent on another more fundamental mimeticaspect: (proto-)imitation.

4.2. Mimesis as iconic representation

A second route to seeking a necessary connection between mimesis and language is therepresentational character of language: the fact that one important function of language isto serve as a “picture of reality” including both iconic (mostly structural) and arbitrary(mostly phonological) features (cf. Itkonen, this volume). But the question is if iconicityis a necessary feature of language, and if mimesis is a necessary road to understandingthe concept of representation. The answer to both questions must be “No”.

The representational function itself could indeed be argued to be necessary – even thoughthis is denied by (strong interpretations of) “pragmatic” theories such as those of the lateWittgenstein (1953) and Austin (1962). Unless a language contains the possibility to beabout some state of affairs, rather than just being “associated” with it, or triggering somebehavior “appropriate” for it, we would hardly call it language. But the mapping betweenthe component parts of the utterance and the represented situation can in principle becompletely arbitrary (even if it is not so in practice), and therefore iconicity can not beconsidered necessary.

It is also impossible to insist that iconicity of the type characteristic for mimesis is anecessary step for understanding the concept of representation, even if this turns out to bethe actual situation in both evolution and ontogeny. For one thing, experiments withteaching symbol systems to apes have shown that they are capable of learning a limitednumber of conventional meanings of symbols, whose form is arbitrary in relation to theirreferents (Savage-Rumbaugh and Rumbaugh 1993, Deacon 1997).

In sum, even if mimesis possesses advantages compared to arbitrary symbol systems interms of transparency and ease of memorization, which motivates its early appearance inboth evolution and ontogeny, these are not strong enough to claim that it is unthinkablethat language could emerge without it.

4.3. Mimesis as a ground for intersubjectivity

A final, and I believe strongest of the three arguments for mimesis playing a necessaryrole for the emergence of language lies in the socially shared nature of linguistic meaning(Wittgenstein 1953, Itkonen 1983, Nelson 1996). This implies not just that themeaning/concept of expression X “inside” Speaker A and Speaker B happens to be thesame, or that A’s and B’s individual concepts becomes somehow “calibrated” with usage(Steels, Kaplan, McIntyre and Van Looveren 2002), but that A and B have an active way

of “negotiating” the meaning of X so that it is the same for both of them, and they knowthat it is the same. The way in which this is done is not too different from the way A andB can determine that they are looking at the same object, i.e. participate in acts of jointattention.

This last comparison suggests where the necessary connection between mimesis andlanguage may lie: joint attention rests on mimesis in so far as it involves the imitation ofpurposeful action. Imitation (in a general sense), whether it rests on a partially innate“mirror neuron system” or not, is very likely the only way for developing structures ofshared experience such as joint attention. In that case, without mimesis there would be nointersubjectivity, and without intersubjectivity, there could be no language. It is hardlyconceivable how language itself would enforce such structures of shared meaning – sincewithout non-linguistic, at least in part mimetic criteria of successful communication,language would be unlearnable.4 In the absence of miraculous factors, mimesis stands asa necessary prerequisite for language. This, I believe, is the meaning behind Donald’ssomewhat cryptic statement:

Language is different from mimesis, but is has mimetic roots. It is acollective product and must have evolved as a group adaptation, in thecontext of mimetic expressive culture. Given the conventional, collectivenature of language, it could not have emerged in any other way. (Donald2001: 274)

In emphasizing that language always occurs in consensual “games” or “forms of life”,Wittgenstein (1953) seems to have made a very similar point. What I am here suggestingis that mimesis is necessary for the establishment of these games, i.e. that they areoriginally “mimetic games”, prior to becoming “language games”. The failure to engagein such games typical for autistic children (Firth 1989) unsurprisingly has detrimentaleffects on language.

5. Conclusions

The discussion in this paper has shown the potential of the concept of mimesis, defied asby Donald (1991, 2001) as conscious, representational body motion, for explaining the“roots” of human language are considerable. In being consistent with a large amount ofevidence from human evolution and ontogeny, it serves a unifying role. Conceptually, itcan help us understand the formation of shared experience, and ultimately sharedsymbolic meaning, which is perhaps the central characteristic – and puzzle – of language:if concepts are psychological entities, how can language be social?

At the same time, one conclusion is to refrain from uncritically applying the concept of a“mimetic stage” as a bridge between the signal-based communication of animals and pre-linguistic children – on the one hand – and the symbolic language of adult human beingson the other. For one thing, mimesis was shown not to be a unitary concept, but to 4 Save for e.g. telepathy.

involve (at least) the phenomena of mime, imitation, skill and gesture some of which areinherently social, others are not, some are explicitly representational, others are not. Anurgent task is to attempt to reconstruct the order in which they have emerged in evolution,which may also parallel their development in childhood. Largely following theVygotskian principle of interpersonal-to-intrapersonal development, a likely progressionis the following:

proto-imitation > imitation > (panto)mime > mimetic schema > gesture

What characterizes this order is that it coincides with increased semioticity, i.e.differentiation of expression from content and an explicit representational relationship,abstractness and internalization. Proto-imitation, both in phylogeny and ontogeny doesnot involve an understanding of the model’s goals, while imitation does. Mime is morerather than less complex, since its function is purely representational rather thanpragmatic. Mimetic schemas are essentially internalized and schematized mimes, andcorrespond structurally to what cognitive linguists call “image schemas” (Johnson 1987)– but a central difference is that since they are based in social rather than individualexperience, they stand in a better position to “ground” linguistic meaning than purelysensorimotor structures. Gestures involve sequences of mimes used for communicativepurposes, differing in degree of conventionality and language-dependence (Kendon1988).

In sum, there is considerable empirical support for the hypothesis that mimesis haspreceded and made possible symbolic language in evolution, and continues to do so inontogeny. But at least in the latter case, this is more in the manner of “waves” serving assupport structures for the development and use of language, rather than a “stages” stableenough to be self-contained. Much of the argumentation presented in this paper has beenconjectural, and there are many questions remaining, but with its potential for filling ingaps in our understanding of semiotic development, for integrating evidence fromphylogeny and ontogeny, and bringing in both empirical and conceptual issues, themimetic hypothesis is certainly worth pursuing further.

References

Acredolo, L. and S. Goodwyn (1996). Baby Signs: How to Talk with your Baby beforeyour Baby Can Talk. Chicago: NTB/Contemporary Publishers.

Acredolo, L. and S. Goodwyn (2000). Baby Minds: Brain-building Games your BabyWill Love. New York: Bantam.

Arbib, M. (in press). Beyond the mirror system: From monkey-like action recognition tohuman language. To appear in BBS.

Armstrong, D. (1998). Review of T. Deacon, “The Symbolic Species: The Co-Evolutionof Language and the Brain”. Evolution of Communication, Vol 2-1: 161-169.

Armstrong, D. Stokoe, W. and S. Wilcox (1995). Gesture, and the Nature of Language.Cambridge: Cambridge University Press.

Austin, J.L. (1962). How to Do Things with Words. Cambridge, Mass.: HarvardUniversity Press.

Bickerton, D. (1990). Language and Species. Chicago: University of Chicago Press.

Birk, L. and A. Winsler (1995). Scaffolding Children’s Knowledge. Vygotsky and EarlyChildhood Education. Washington D.C.: National Foundation for the Education ofYoung Children.

Butterworth, G. and N. Jarrett (1991). What minds have in common is space: spacemechanisms serving joint visual attention in infancy. British Journal ofDevelopmental Psychology 9: 55-72.

Corballis, M. (2002). Did language evolve from manual gestures. In A. Wray (ed.) TheTransition to Language. Oxford: Oxford University Press.

Corballis, M. (1991). The Lopsided Ape: Evolution of the Generative Mind. Oxford:Oxford University Press.

Deacon, T. (1997). The Symbolic Species: The Co-Evolution of Language and the Brain.NY: Norton.

Donald, M. (1991). Origins of the Modern Mind. Three Stages in the Evolution of Cultureand Cognition. Cambridge, Mass.: Harvard University Press.

Donald, M. (2001). A Mind So Rare. The Evolution of Human Consciousness. New York:Norton.

Evans, G. and K. Rubin (1979). Hand gestures as a communicative mode in school-agedchildren. Journal of Genetic Psychology, 135: 189-196.

Frith, U. (1989). Autism: Explaining the Enigma. Oxford: Basil Blackwell.

Givón, T. (1998). On the co-evolution of language, mind and brain. Evolution ofCommunication, Vol 2-1: 45-116.

Greenfield. P. (1991). Language, tools and brain: the ontogeny and phylogeny ofhierarchically organized behavior. Behavioral and Brain Sciences, 14: 531-595.

Grice, P. (1957). Meaning. Philosophical Review 66: 377-88.

Gullberg, M. (1998). Gesture as a Communication Strategy in Second LanguageDiscourse. A Study of Learners of French and Swedish. Lund: Lund UniversityPress.

Haeckel, E. (1874). The Evolution of Man: A Popular Exposition of the Principle Pointsof Human Ontogeny and Phylogeny. New York: International Science Library.

Haiman, J. (1985). Natural Syntax: Iconicity and Erosion. Cambridge: CambridgeUniversity Press.

Itkonen, E. (1983). Causality in Linguistic Theory. A Critical Investigation into thePhilosophical and Methodological Foundations of 'Non-autonomous' Linguistics.Bloomington: Indiana University Press.

Johansson, S. (2002). The Evolution of the Human Language Capacity. MA Thesis,Department of Linguistics, Lund University.http://www.ling.lu.se/education/essays/SverkerJohansson_D.pdf

Johnson, M. (1987). The Body in the Mind. Chicago: University of Chicago Press.

Kendon, A. (1988). How gestures can become like words. In F. Poyatos (ed.)Crosscultural Perspectives in Nonverbal Communication. Toronto: Gogrefe.

Klein, N. (2000). No Logo. London: Harper Collins.

Lakoff, G. (1987). Women, Fire and Dangerous Things: What Categories Reveal Aboutthe Mind. Chicago: University of Chicago Press.

Lieberman, P. (1991). Uniquely Human: The Evolution of Speech, Thought and SelflessBehavior. Cambridge, MA: Harvard University Press.

McNeill, D. (1992). Hand and Mind. What Gestures Reveal about Thought. Chicago: TheUniversity of Chicago Press.

McNeill, D. (2000). Analogic/analytic representations and cross-linguistic differences inthinking for speaking. Cognitive Linguistics: Vol. 11, 1/2: 43-60.

Meltzoff, A. and A. Goopnik (1993). The role of imitation in understanding persons anddeveloping a theory of mind. In S. Baron-Cohen, H. Tager-Flusberg and D. Cohen(eds.) Understanding Other Minds. Perspectives from Autism. Oxford: OxfordUniversity Press.

Meltzoff, A. and M. Moore (1977). Imitation of facial and manual gestures by humanneonates. Science 198: 75-78.

Meltzoff, A. and M. Moore (1983). Newborn infants imitate adult facial abilities. ChildDevelopment 54 (3): 702-9.

Meltzoff, A. and M. Moore (1995). Infants’ understanding of people and things: Frombody imitation to folk psychology. In J. L. Bermudez, A. Marcel, and N. Eilan(Eds.), The Body and the Self. Cambridge, Mass.: The MIT Press.

Morford, J., Singleton, J. and S. Goldwin-Meadow (1995). The genesis of language: Howmuch time is needed to generate arbitrary symbols in a sign system. In K. Emmoreyand J. Reilley (Eds.) Language, Gesture, and Space. Lawrence Erlbaum.

Müller, R. (1996). Innateness, Autonomy, Universality? Neurological Approaches toLanguage, Behavioral and Brain Sciences 19, 611-675.

Nelson, K. (1996). Language in Cognitive Development. The Emergence of the MediatedMind. Cambridge: Cambridge University Press.

Piaget, J. (1952). Play, Dreams and Imitation in Childhood. London: Routledge andKegan Paul.

Piaget, J. (1953). The Origin of Intelligence in the Child. London: Routledge and KeganPaul.

Piaget, J. (1971). Biology and Knowledge. Chicago: University of Chicago Press.

Piattelli-Palmarini, M. (1989). Evolution, selection and cognition: from ”learning” toparameter setting in biology and in the study of language. Cognition 31: 1-44.

Pierce, C. S. (1931-35). The Collected Papers of Charles Sanders Peirce. Vols. 1-4.Cambridge, Mass.: Harvard University Press.

Premack, D. (1987). Gavagai. Cambridge, MA. MIT Press.

Rizzolatti, G. and M. Arbib (1998). Language within our grasp. Trends in Neurosciences.21: 188-194.

Saussure, F. de (1916). Cours de Linguistique Générale. Paris: Payot.

Savage-Rumbaugh, S. and Rumbaugh, D. (1993). The emergence of language. In K.R.Gibson and T. Ingold (Eds) Tools, Language and Cognition in Human Evolution.Cambridge: Cambridge University Press.

Sinha, C. (1988). Language and Representation. A Socio-naturalistic Approach toHuman Development. New York: Harverster Press.

Sinha, C. (2001). The epigenesis of symbolization. In C. Balkenius, J. Zlatev, H. Kozima,K. Dautenhahn and C. Breazeal, (eds.) Proceedings of the First InternationalWorkshop on Epigenetic Robotics, Lund University Cognitive Science, 85.

Steels, L., Kaplan, F., McIntyre, A. and J. Van Looveren (2002). Crucial factors in theorigins of word-meaning. In A. Wray (ed.) The Transition to Language. Oxford:Oxford University Press.

Stokoe, W. (2001). Language in Hand: Why Sign Came before Speech. Washington DC:Gallaudet University Press.

Tomasello, M. (1995). On the interpersonal origins of the self-concept. In U. Neisser(Ed.) The Perceived Self. Ecological and Interpersonal Sources of Self-knowledge.Cambridge: Cambridge University Press.

Tomasello, M. (1999). The Cultural Origins of Human Cognition. Cambridge, MA:Harvard University Press.

Trevarthen, C. (1992). An infant’s motives for speaking and thinking in the culture. In A.Wold (Ed.) The Dialogical Alternative, Towards a Theory of Language and Mind.Oslo: Scandinavian University Press.

Vygotsky, L. (1962). Thought and Language. Cambridge, Mass.: MIT Press.

Vygotsky, L. (1978). Mind in Society. The Development of Higher PsychologicalProcesses. Cambridge, Mass.: Harvard University Press.

Waddington, C.H. (1975). The Evolution of an Evolutionist. Edinburgh: EdinburghUniversity Press.

Wittgenstein, L. (1953). Philosophical Investigations. Oxford: Basil Blackwell.

Zlatev, J. (1997). Situated Embodiment, Studies in the Emergence of Spatial Meaning.Stockholm: Gotab Press.

Zlatev, J. (2000). The mimetic origins of self-consciousness - in phylo-, onto- androbotogenesis. Proceedings of the 26th Annual Conference of the IEEE IndustrialElectronics Society, October 2000, Nagoya.

Zlatev, J. (2001a). The epigenesis of meaning in human beings, and possibly in robots.Minds and Machines, 11 (2): 155-195.

Zlatev, J. (2001b) A hierarchy of meaning systems based on value. In C. Balkenius, J.Zlatev, H. Kozima, K. Dautenhahn and C. Breazeal, (eds.) Proceedings of the FirstInternational Workshop on Epigenetic Robotics, Lund University CognitiveScience, 85.